Bart Smaalders wrote:
> Garrett D'Amore wrote:
>> Jyri Virkki wrote:
>>> Garrett D'Amore wrote:
>>>
>>>> Ah, but the same logic is used to cite "inclusion" of other FOSS
>>>> packages. For example, unison does not ship in the default Ubuntu
>>>> distribution, but is part of "Universe" (which is fairly similar to
>>>> the FreeBSD ports collection.)
>>>>
>>>> I wonder how much of what we are including in our "base" is stuff
>>>> that really belongs in a readily accessible network repository.
>>>
>>> As we evolve away from consolidations and WOS it's important to be
>>> clear with terminology. Not sure what you mean by 'default
>>> distribution'. Plenty of Sun's very own packages are not going to be
>>> on the LiveCD (possibly the most obvious interpretation of 'default
>>> distribution') and live only on a readily accessible network
>>> repository to be installed at will, or not.
>>>
>>
>> Right now, though, the only metric we have is the "Everything Plus
>> OEM" installation in OEM. That is, effectively, the "default
>> install" for Solaris (and OpenSolaris) customers today. (One could
>> argue that the "Developer" meta cluster fills this role instead. I
>> won't debate the matter one way or another...)
>>
>> Once we have a network repo, and a smaller LiveCD (with hopefully
>> more care thought put into what packages are part of the LiveCD) a
>> lot of my objections to past FOSS cases will simply vanish. (I'm
>> sure "unzoo" is useful to someone somewhere, and I don't want to deny
>> him access to it. I just don't want to saddle everyone else with
>> baggage that serves only a tiny minority of users, and which fails to
>> meet certain minimum architectural and possibly quality -- or at
>> least quality assurance -- standards.)
>
> Perhaps you could simply accept that we're going to have a networked
> repository, rather than raising the same pointless issue on every FOSS
> case that comes up. We simply cannot wait to begin integrating more
> FOSS into Nevada until we switch packaging systems.
Does this "cannot wait" attribute apply to every bit of FOSS out there?
I certainly agree that some FOSS is widely useful, and not having it in
Solaris represents a potential feature disparity that we need to close.
However, I remain unconvinced that this is true of all of the FOSS cases
we have been presented with.
My personal preference would be to "slow" the rate of FOSS inclusion, so
that the software that is included is limited to stuff that we think
will have widespread appeal and which either comes already suitable for
Solaris (i.e. no major architectural flaws, and hopefully suitably high
quality as verified by QA) or which we make suitable by appropriate
modifications.
Once a real network repo is available with support for some kind of
tagging, then the floodgates for all the rest of the FOSS software (of
whatever architecture or quality) could be opened up.
But as I mention below, at this point, I'm going to stand aside and let
others make what decisions they will.
>
> I don't see what your objections are, actually. Are you saving space
> on the
> DVD for inclusions you see as more valuable? As has been pointed out
> before, your
> perception of FOSS as inherently lower quality is a stereotype, and
> ignores
> the origin and status of much of the software Sun ships. The majority
> of the software in Solaris is Open Source, and some of our closed source
> has not been enhanced in years.
Much of the FOSS we have historically shipped has been "customized" to
fit our particular needs, so that it conforms well with Solaris. For
example, the very large amount of effort spent on JDS/Gnome, Sendmail,
and even perl. Even utilities like tcsh and bash behave, mostly, like
native Solaris applications at least.
My concerns grow when two things happen:
1) the cost of carrying the software to every Sun/Solaris customer is
likely to be significantly higher than than the benefits derived by
inclusion (unzoo is a good case in point here) -- bits on the DVD don't
come for free -- and sustaining support for FOSS isn't free either!
2) the product falls far short of the architectural expectations that we
would apply for Sun-derived source (whether Open Source or not) -- I
think Unison's large architectural gaps fit this bill
I do believe that there is significant value in a system of
well-architected and tailored (or at least "appearing to be so" by
working well together) software that fits the definition of what I would
call a "core platform". This is a value that Solaris has (IMO)
traditionally offered, and is one of the main distinguishing benefits
(and some would also say detriments) between Solaris and many of the
Linux distros.
Perhaps those old values are outmoded, and our best way forward really
is to integrate more FOSS than Linux. Certainly it seems like the Web
2.0 market largely just wants to have everything ever developed, and the
most recent versions, available right away, and maybe that market is now
more important to Solaris than the Enterprise market where stability and
fit-and-finish are perhaps more valued.
>
> The next version of Solaris will ship w/ a networked repository, and no
> consistent definition of "default" install.
Since there is no "next version of Solaris" yet on the roadmap (at least
not publicly), its hard to know what it will, or will not, ship with.
> With an ever increasing set
> of available packages, very few people will want to "install everything";
> in fact, we anticipate providing no option to do so.
Well, many people don't like to make choices, but they don't want to
make a choice that might exclude functionality they want. So if you
don't have an "everything", but instead have a "core + developer + some
other stuff" that will be what is chosen.
But in the absence of a clear public plan here, I can only presume based
on the architecture we have *today*.
Okay, with that all said, I'm going to back down completely from the
FOSS cases. It seems like maybe my viewpoint is not terribly popular,
and raising the same arguments over and over again is just getting in
the way of "progress" (both for me and for the FOSS project teams).
(And at this point everyone who needs to has already heard my concerns
and generally brushed them aside.) If any other ARC members feel the
same as I do, and care enough to say so, then let them be heard.
Otherwise, let the FOSS smorgasborg commence.
-- Garrett
